Fingerprints in the Air: Unique Identification of Wireless Devices Using RF RSS Fingerprints
Fingerprints in the Air: Unique Identification of Wireless Devices Using RF RSS Fingerprints
复制标题
空中指纹：使用 RF RSS 指纹对无线设备进行唯一识别
DOI：
10.1109/jsen.2017.2685564
复制
发表时间：
2017-06-01
影响因子：
4.3
通讯作者：
Liu, Zhi
中科院分区：
文献类型：
--
作者：
Li, Qiyue;Fan, Hailong;Liu, Zhi
近年来，室内定位、设备识别和无线考勤安全系统蓬勃发展。这些解决方案通常利用RF接收信号强度指纹来定位人员。然而，它们带来了一个重要但通常被忽视的问题，即检测一个人是否携带了多个无线设备。换句话说，应该将其他设备排除在分析之外。为了检测唯一的身份识别问题，在入口处附近部署了生物辅助方法，如指纹、人脸和步态识别。然而，这些方法不仅难以实施，而且还会带来额外的成本。本文研究了射频接收信号强度指纹的唯一性识别问题，采集的信号强度指纹被建模为时间序列。通过计算时间序列的相似度来实现唯一识别。具体地，提出了一种基于动态时间规整的朴素算法来计算异步时间序列的相似度。在此基础上，提出了一种基于特征提取和谱聚类的两步改进算法，降低了相似性检验的计算复杂度。此外，还提出了一个有效性指数来确定最优聚类个数。仿真和实验结果表明，在典型场景下，我们的算法可以检测到计算复杂度适中的唯一识别问题。
Recent years have witnessed the proliferation of indoor localization, device identification, and wireless attendance security systems. These solutions typically leverage RF received signal strength fingerprints to locate persons. However, they entail an important albeit commonly ignored issue, i.e., detecting whether an individual is carrying more than one wireless device. In other words, additional devices should be excluded from the analysis. To detect the unique identification problem, bio-assisted methods, such as fingerprint, face, and gait recognition, are deployed near entrances. However, these methods are not only difficult to implement but also entail additional costs. This paper studies the unique identification problem using RF received signal strength fingerprints, which are collected and modeled as time series. The similarity of the time series is calculated to achieve unique identification. Specifically, a naive algorithm based on dynamic time warping is proposed to compute the similarity in the asynchronous time series. Then, an improved two-step algorithm based on feature extraction and spectral clustering is proposed to reduce the computational complexity of the similarity check. In addition, an effectiveness index is proposed to obtain the optimal number of clusters. The results of simulations and experiments show that our algorithms can detect the unique identification problem with moderate computational complexity in typical scenarios.